Migration and Settlement: 3. Sweden

MIGRATION AND SETTLEMENT:

3. SWEDEN

h e E. Andersson

International Institute for Applied Systems Analysis

Ingvar Holmberg

University o f Gothenburg

RR-80-5 March 1980

INTERNATIONAL INSTITUTE FOR APPLIED SYSTEMS ANALYSIS Laxenburg, Austria

Research Reports, which record research conducted at IIASA, are independently reviewed before publication. However, the views and opinions they express are not necessarily those of the Institute or the National Member Organizations that support it.

Copyright O 1980

International Institute for Applied Systems Analysis

AU rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage or retrieval system, without permission in writing from the publisher.

FOREWORD

Interest in human settlement systems and policies has been a central part of urban-related work at the International Institute for Applied Systems Analysis (IIASA) from the outset. From 1975 through 1978 this interest was manifested in the work of the Migration and Settlement Task, which was formally concluded in November 1978. Since then, attention has turned to dissemination of the Task's results and to the conclusion of its comparative study, which, under the leadership of Dr. Frans Willekens, is focusing on a comparative quantitative assessment of recent migration patterns and spatial population dynamics in all of IIASA's 17 National Member Organization countries.

The comparative analysis of national patterns of interregional migration and spatial population growth is being carried out by an international network of scholars who are using methodology and computer programs developed at IIASA.

This report reviews recent multiregional population changes in Sweden.

Professor Ake Anderson of the University of UmeP, Sweden, on leave at IIASA, and Associate Professor Ingvar Holmberg, of the University of Gothenburg, ana- lyze the demographic components of spatial population change and evaluate the effects of population distribution policies implemented over the past decades.

Reports summarizing previous work on migration and settlement at IIASA are listed at the end of this report.

Andrei Rogers Chairman Human Settlements

and Services Area

CONTENTS

1 INTRODUCTION

1.1 Purpose and Background

1.2 The Administrative Subdivisions of Sweden 1.3 Organization of Swedish Population Statistics

1.4 The Historical Settlement Pattern of Sweden 1750- 1975 2 CURRENT PATTERNS OF SPATIAL POPULATION GROWTH

2.1 Regional Population Development

2.2 Regional Disaggregation and Aggregation used in the IIASA Population Projections

2.3 Observations of Aggregation Errors Due to Limitation of the Number of Regions

3 MULTIREGIONAL POPULATION ANALYSIS 3.1 Introduction

3.2 The Multiregional Life Table 3.3 Fertility and Mobility Analysis

3.4 The Multiregional Population Projection

3.5 On Fluctuations in Swedish Demographic Behavior 4 REGIONAL POPULATION POLICIES IN SWEDEN

4.1 An Historical Outlook

4.2 From Efficiency-Oriented to Equity-Oriented Regional Population Policy

4.3 Evaluation of Regional Policy Effects 5 CONCLUSIONS

REFERENCES

APPENDIXES

A Observed Population and Numbers of Births, Deaths, and

Migrants by Sex, Age, and Region, 1974 62

B

Total Population, 1974 7 6

C Expectations of Life by Region of Birth and Region of Residence,

Total Population, 1 974 88

D Multiregional Population Projection (Constant Age-Specific Rates)

8 Regions, 1974-Stability 9 8

1 INTRODUCTION

1.1 Purpose and Background

This report is part of the Comparative Migration and Settlement Study included in the Migration and Settlement Task in the Human Settlements and Services Area at the International Institute for Applied Systems Analysis (IIASA). The purpose of this report is t o give a detailed overview of the internal migration patterns and regional policies in Sweden. As similar studies are being carried out in all the member countries of IIASA, a basis is being laid for comparing migration patterns.

Since the technique of multiregional population analysis is employed in the calculations, an additional by-product of the study is an evaluation of this method and its usefulness as a tool for policy makers in the study of human settlement systems and for regional population forecasts.

Sweden has detailed population data dating from the middle of the eigh- teenth century. In 1748 an organization for the collection and tabulation of population statistics was established which has made it possible t o study popula- tion development and its components over a long time perspective.

In the 1970s Sweden had a program for collecting vital statistics and other population data which allowed excellent opportunities for evaluating new tech- niques of population analysis. This program has few counterparts in other coun- tries with respect t o its completeness, coverage, and exactness.

This report is organized as follows. Sections 1.2 and 1.3 discuss the admin- istrative subdivisions of Sweden and the organization of the population statistics, respectively. Section 1.4 describes the settlement pattern of Sweden in a histor- ical perspective. Chapter 2 deals with current patterns of spatial population growth and of its components (fertility, mortality, and migration). The problem of regional aggregation is also discussed, t o some extent, and various regional

systems are proposed. In Chapter 3 the results of the multiregional population analysis are presented: life tables, population projections, and fertility and migration analysis. Chapter 4 presents an outline of population policies in Sweden, with special emphasis on the regional labor market and internal migra- tion policies.

1.2 The Administrative Subdivisions o f Sweden

The basic unit in the population registration system of Sweden is the parish.

This can be subdivided into one or more districts. For administrative purposes the basic unit is a municipality (kommun), and with approximately 2,570 parishes and 278 municipalities there are, on average, 9 or 10 parishes in each municipality. The municipalities may be further aggregated into 7 0 A-regions -

as they are known - or t o 24 counties (Figures 1.1 and 1.2).

The A-regions were introduced, as part of the labor market policy organiza- tion, at the end of the 1950s. They are defined, for the purpose of labor and service administration, as "commuting regions." When presenting population, economic, o r other statistics the counties are often aggregated into county regions, of which there are eight. The boundaries of these county regions are shown in Figure 1.1

.

The population density of the A-regions varies greatly between different parts of the country. For this reason most of the northern A-regions are not really commuting regions but rather public service areas and labor-market plan- ning regions. The importance of this distinction is shown in Table 1.1.

The parish is the basic unit in the population registration system and it may also be used as a basic unit in demographic analysis. However, such a large number of regions cannot be handled for most practical purposes. The regional system actually used depends on the purpose of the study. In the historical out- line of Swedish population growth the county regions have been used, mainly because these regions give an acceptably accurate picture of the population development in various parts of the country. More recent development trends are also illustrated by the A-regions. Other regional systems are discussed in section 2.1 and may be regarded as special-purpose regions.

1.3 Organization o f Swedish Population Statistics

According t o Swedish law everyone who permanently lives in Sweden should be registered in one of the parishes of the country. The parish registers are kept by the clergy and a register is also kept at each of the 24 counties in the country.

With only a few exceptions all vital events, such as births, deaths, changes in marital status, and external and internal migratory moves, are reported t o the parish and by the parish to the County Board. Any changes are sent weekly t o the County Board where the information is used for updating the Register of the Total Population (RTB) which is kept at the Central Bureau of Statistics in 0rebro.

Wes

(lo

Jpper North

Name of county AB Stockholm C Uppsala D Sodermanland E t)stergotland F Jonkoping G Kronoberg H Kalmar I Gotland K Blekinge L Kristianstad M Malmohus N Halland 0 Goteborgs och P Alvsborg R Skaraborg S Varmland T drebro U Vastmanland W Kopparberg X Gavleborg Y Vasternorrland Z Jamtland AC Vasterbotten BD Norrbotten

Bohus

FIGURE 1 . 1 Counties and county regions. Source: Central Bureau of Statistics.

01 StockholmlSb'dertalje 02 Norrtalje, 03 Enkoping 04 Uppsala. 05 Nykoping 06 Katrineholm, 07 Eskilstuna 08 Miolby IMotala

09 Linkoping, 10 Norrkoping 1 1 Jonkoping, 12 Tranas 13 EksjolNassiolVetlanda 14 Varnamo, 15 Ljungby 16 Vaxjo, 17 Vastewi k 18 HultsfredIVimmerby 19 Oskarshamn 20 KalmarINybro 21 Visby, 22 Karlskrona 23 Karlshamn, 24 Kristianstad 25 Hassleholm, 26 Angelholm 27 HelsingborgILandskrona 28 Malmo/Lund/Trelleborg 29 YstadISimrishamn 30 Eslov, 31 Halmstad 32 FalkenbergIVarberg 33 ~oteborg/Alings%s 34 Uddevalla

35 TrollhattanIVanersborg 36 Bor%s

37 LidkopinglSkara 38 Falkoping, 39 Skovde 40 Mariestad

41 KristinehamnIFiIipstad 42 Karlstad

43 ~ ~ f i f f ~ e / A m P ~ 44 Arvika, 45 Urebro 46 Karlskoga 47 Lindesberg

48 ~ L t e r a s , 49 Koping 50 Fagersta, 51 Sala 52 BorlangeIFalun 53 AvestaIHedemora 54 Ludvika, 55 Mora 56 GgLleISandviken 57' BollnhlSoderhamn 58 Hudiksvall/Ljusdal 59 Sundsvall

60 Harnosand/Kramfors 61 Sollefte%

62 drnskoldsvik 63 Ostersund 64 Ume%

65 Skelleftei 66 Lycksele 67 PiteD 68 Lule%/Boden 69 HaparandaIKalix 70 KirunaIGallivare

FIGURE 1.2 A-regions. Source: Labor Market Board

TABLE 1.1 Average area and population of A-regions belonging to different county regions in Sweden (1 974).

County region Stockholm East Middle South Middle South West North Middle Lower North Upper North

Population Area in 1,000 km2 3,247

2,745 2,974 1,541 2,989 5,344 14,326 22,048

SOURCE: Folk- och bostadsr&ningen 1975. De13.2 (Population Census 1975. Part 3.2).

NOTE: The A-region boundaries do not coincide exactly with the boundaries o f county regions, which causes a minor error in the figures.

In order to have a civil registration system where all increases and de- creases balance at zero a special register called the Residence Unknown Register has been introduced. Until 1967 all persons whose place of residence was un- known at two successive annual registration controls were transferred t o this register. In 1968 the procedure was changed slightly and foreign citizens who cannot be assumed to live in Sweden anymore and Swedish citizens who at two successive annual registrations have a permanent residence abroad are recorded as emigrants. In most cases transfers from the Residence Unknown Register concern persons who have returned to Sweden but who never reported their original emigration.

When regional comparisons of demographic characteristics are made it is important t o know the principles behind the registration of vital events. New- born children are registered in the parish where the mother was registered at the time of delivery. The only exceptions t o this are children born t o women in the Residence Unknown Register or to foreigners who are not registered in the country at all. In the latter cases the birth will be registered in the parish where the delivery takes place. Deaths are recorded in the parish where the deceased was registered or, if not registered at all, in the parish where death took place or the corpse was found.

In population statistics migrations are defined as moves across parish borders. The time of migration is defined as the week when the move is recorded at the County Board. In the case of an internal migration over a county border the move is recorded both in the county of origin and in the county of destina- tion, and the time of the move refers to week of registration in the county of origin.

It is important to know something about the effect of the "cut-off weekv.*

*The sixteenth week of the year.

TABLE 1.2 Registrations of vital events by year of occurrence (1 976).

Year of occurrence Births Deaths

Total 5 3 75

SOURCE: Befokningsforandringar 1976. Del 3 (Population Changes.

Part 3).

Since all vital statistics are based on announcements from the parish, a certain time lag appears in the registration of changes. The present procedure includes all information up t o the sixteenth week, of the year following the statistical year. The general population "cut-off week" is the fourth week following the end of the statistical year. Because o f this a certain, but small, discrepancy exists with respect t o population and vital events reported for any given year.

Table 1.2 gives an idea of the size of this error. The total number of births amounts t o 353. If the distribution over a year of occurrence is reasonably stable the total error in the birth statistics could be estimated at approximately 3.6 per thousand. For the more recent years of occurrence it is probably a question of the delay in reporting events that causes the error. F o r earlier years of occur- rence it may have been caused by parents returning t o their home country and reporting births that took place abroad. By the same assumption the error in the death statistics could be estimated at less than 1 per thousand. In this case some o f the oldest registrations are official declarations of death for persons who have disappeared.

There is also a slight error in the total population reported at the end of the year. As an earlier "cut-off week" is employed in population statistics than in statistics of vital events there is a small deviation between the difference in populations at the end and the beginning of the year, and the total number of changes reported for the same year.

An extensive check of the civil registration was carried out in 1972 when Sweden changed t o a new system for car registration. Car registration cards were sent to all persons o f ages 15-90. In all cases of nondelivery the reasons were investigated by the parish authorities. The total number of undeliverable cards was 64,700. The causes are summarized in Table 1.3.

TABLE 1.3 Results of the civil registration system check (1972).

Undeliverable register cardsO Cause of nondelivery

Total number of undeliverable register cards Move reported before investigation

No measures taken incomplete address deceased

temporarily absent

Not living at given address but should be recorded in the parish register (e.g., hospitalized for long-term care) Migrated without notifying authorities

migrated within the country migrated to other Nordic country migrated to non-Nordic country Residence unknown

citizen of other Nordic country citizen of non-Nordic country other

' ~ i ~ u r e s within parentheses are percentages of the total population aged 15-90 years.

SOURCE: Befolkningsforiindringar 1973. De13 (Population Changes 1973. Part 3).

Depending on the definition there is a total maximum error of 0.4 percent and a minimum error of 0.1 percent in the population figures for ages 15-90.

1.4 The Historical Settlement Pattern o f Sweden 1750-1975

The Swedish population distribution is available since the beginning of the eighteenth century. Population data from 1750 have been aggregated into the 8 county regions and are illustrated in Figure 1.3.

On the basis of Figure 1.3 the following observations about the settlement pattern in Sweden can be made:

- the Stockholm region has been increasing its relative share of the total population at a steady rate from 1850 t o 1970. This was broken by a period of stagnation beginning in about 1970

- the counties of the periphery (Upper and Lower North) have also shown an increase. From a share of approximately 5 percent around 1750, they increased t o 8 percent around 1850 and had a peak of

-

^-

a C Stockholm

2 a

0 East Middle

- _.- 5

C m South

-

-

3 ₀ 10-

c South Middle

w- 0 a L

z m

V] Upper North

5

-

I 1 I I I

1 800 1900 1970 Year

FIGURE 1.3 Population distribution in Sweden 1750-1974. Source: 1750-1950 Historisk Statistik I. Befolkning (Historical Statistics I. Population); 1951-1960 Befollcningsrorelsen (Population Changes); 1961 -1 966 Folkmingdens for'5ndringar (Population Changes); 1967- 1974 Befolkningsfijr'indringar. Del 3 (Population Changes. Part 3).

approximately 13 percent in 1950. Since then their relative share has decreased and in 1970 it was slightly less than 11 percent of Sweden's total population

- the most stable parts of the country are the southern and the western regions which have had approximately 14 and 19 percent of the total population, respectively, for the time period illustrated

- all the counties of the eastern and mid-inland parts of the country (South Middle, East Middle, and North Middle) have declined in rela- tive economic and demographic importance. These areas had around 53 percent of the total population in 1750, a figure that had decreased t o 4 4 percent by 1900 and t o 37 percent by 1970

The long-term migration movement appears t o be from the inland to the coastal areas, with some concentration in the metropolitan region of Stockholm.

In a historical perspective there has been no real increase in the relative popula- tion shares of the Gothenburg and Malmo metropolitan regions. The development

of the settlement pattern has been accompanied by significant changes in popu- lation densities and in the regional income distribution. The development of the population density in the county regions isgiven in Figure 1.4 and the change in income shares for the period 1920- 1975 is shown in Figure 1.5.

Changes in the settlement pattern cannot be illustrated only with statistics for macro regions, of the kind used in the study. Large pattern changes can only be recorded at lower levels of aggregation. This is particularly the case for devel- opment after 1950, when the great revolution in transportation and communi- cation technology occurred. The introduction of mopeds, motorcycles, and cars as private means of transportation and the expansion of the telephone, television, and other electronic networks have made it less important t o have a central loca- tion in the public transportation networks.

To illustrate these changes two main levels of geographical aggregation are relevant - the labor-market regions and the urbanlrural regions. The first is related to the A-regions mentioned previously. Their average population is 120,000 but variation in size is large and follows the well-known rank-size rule fairly accurately. The development of the rank-size rule for these A-regions is shown in Figure 1.6 for the period 1965-1975.

It has been argued that regional differences in population growth have become less pronounced. This implies an equalization of the regional distribu- tion of the country's population. This means that for rank-size distribution the curve should be displaced upward, showing a slower descent. Figure 1.6 shows that this is not true and that, in fact, the descent has become much steeper, im- plying an even more pronounced inequality of population distribution. The changes in the ranks over the 10-year period are, in most cases, almost insignif- icant. There are a few exceptions, however, pertaining t o either growing A-regions near t o the three metropolitan regions, e.g., Nykoping, ;ingelholm, and Enkoping (Figure 1.6), or declining A-regions in sparsely populated areas, such as Ly cksele which has dropped from rank 4 9 to rank 58. Although the rank-size rule has been regarded with great suspicion by many scholars, it is interesting to note how well this rule describes the regional distribution of the population in Sweden.

It also indicates that the suggestion that the population distribution in Sweden has become more even during recent years is not well founded.

A characteristic problem observed in most industrialized countries is the great exodus of people from rural t o urban areas. Although statistics are scarce for earlier periods it is possible t o study the development of the urban popula- tion from the beginning of the nineteenth century, and some rough estimates can be made back t o the mid-eighteenth century. Even today large areas in Sweden are very sparsely populated. An important change in nineteenth-century Sweden was the population increase in the more peripheral areas. The coloniza- tion of large parts of northern Sweden played an important part in this popula- tion redistribution. From 1880 onward the population increase became more and more concentrated around certain growth poles, where the population increased more rapidly than it did for the country as a whole.

I

Year

FIGURE 1.4 Density of population in the county regions. Source: Population: see Figure 1.3. Area: 1880-1 950 FolkrPkningen (Population Census); 1960-1975 Statistisk Arsbok (Statistical Abstract of Sweden).

Wen East Middle

E

North Middle South Middle Upper North Lower North

v

1920

Year

FIGURE 1.5 Income shares for regions as a percentage of the total income of Sweden.

Source: Skattetaxeringarna 1922-1971 (Income assessment);Taxeringsutfallet 1975. Statis- tiska meddelanden N 1975 :92 (Tax assessment 1975. Statistical Reports).

NOTE: In Sweden this concept of income is defined as total personal income less deductions for deficit in income source and other general deductions.

t

1 I I I I I I I I I I 1 I

1 10 20 30 40 50 60 70

Rank 1000-

900

-:

800

-

700-.

d 5 _VI 600-

5

z

.- C

c,

.- _I

-

FIGURE 1.6 Rank-size distribution o f Swedish A-regions. Source: Calculated from raw data. 1965 Folk- och bostadsiakningen. De13 (Population Census. Part 3); 1975 Folkmangd.

Del3 (Population Census. Part 3).

I I I i ,

I I I I

1

, Goteborg

.

1 I

I

:

I

a 0 n

+ Population 1965 o Population 1975

-

Lycksele

TABLE 1.4 Number of localities by size of population; 1950, 1960, 1970, and 1975.

Size of population

Number of localities

Total 2,056 1,814 1,775 1,786 87

umber of localities in 1975 as a percentage of the number in 1950.

The major reason for this localized growth is the emergence of cities. For the mid-eighteenth century the total urban population has been estimated at approximately 9 percent of the total population. In 1880, when actual data are available for the fnst time, the urban share of the total population had not reached 10 percent. It was not until the middle of the nineteenth century that the urban share of the total population exceeded 10 percent. From then the urban population increased at a growing rate. At the end of the nineteenth cen- tury the share had doubled and in the 1970s approximately 80 percent of the population lived in "localities," especially around Stockholm and along the west coast.

Localities are here defined as agglomerations with more than 200 persons in a contiguous location. Contiguity implies that n o people who live more than 200 m from their nearest neighbor are included in the locality. Some exceptions t o this rule can be seen, especially in the larger localities where interactions have been observed t o operate over distances larger than 200 m. When a large proportion of the population lives in localities it is implied that the internal size distribution of population localities is of importance. The evolution of the rela- tive importance of localities of different size is shown in Table 1.4.

The number of localities with less than 2,000 inhabitants has decreased.

The most striking decrease has been in the smallest size group. All the size classes above and including 2,000 inhabitants have increased rather rapidly, with a peak in the increase of the localities of 50,000-100,000 inhabitants. In addition t o this, an important concentration of the population in certain regions has become evident, (see Figure 1.7). The natural increase has shown large regional varia- tions, but it is primarily internal migration that has contributed t o the more rapid growth of certain regions.

Parallel with the redistribution of population and the urbanization of the

storstader primara centra

FIGURE 1.7 Local population in 1975 (in thousands). Source: Studies in Regional Policy (in Swedish) Labor Market Department, Ds A 1975 : 12.

NOTE: Local population is defined as the number of people within circles of radius 30 km centered on 1,200 nodes that are spread evenly over the country.

country, a restructuring of the population has taken place with respect t o social class and occupation. The agricultural population, which for a long time was com- pletely dominant, reached its maximum size in about 1880, when it amounted t o approximately 3 million. Since then a continuous decline has taken place and the agricultural proportion of the population was only about 5 percent in the 1970s. In 1940 the population working in manufacturing industries had become the largest population group in the country, accounting for 38 percent of the total population. This proportion is now decreasing and the population in the service sector is the most rapidly increasing part of the working popula- tion.

The historical regional population and economic development has been extensively analyzed in several official and semi-official reports in the SOU*

and ERU*

*

2 CURRENT PATTERNS O F SPATIAL POPULATION GROWTH 2. I Regional Population Development

Data on components of regional population development are only available t o a limited extent. This is due to the organization of the population statistics.

Age-distribution data on births and deaths, as well as total net internal and total external migration, are given for each year over the last 25-year period on a regional basis. Corresponding population data, however, are only given at the population censuses. Since the revision of population statistics in 1967 popula- tion data with regional disaggregations have become available. As a consequence, the presentation of components of regional population change often has t o be based on crude rates although more detailed calculations, centered around the census years, may give a fairly representative view of the development.

Figure 2.1 shows the contribution of various components t o total regional population growth. The rapid urbanization of the 1950s and 1960s is represented by the growth of the metropolitan regions (Stockholm, South, and West). When seen in this perspective Lower North seems t o be the most problematic region, showing an almost constant decrease in population. The out-migration from this region is not compensated for by a natural increase or net external migra- tion surplus as is the case for the Upper North region. In the latter region natural increase is larger than the net internal out-migration up t o 196 1 when the region experienced a population loss for a few years. The same observations can be made for the North Middle region. In the South Middle region a balance exists between natural increase and net internal migration losses. When a region shows a more rapid increase in the 1960s this is mainly due t o net external migration gains a t that time.

The general trend in regional population development was broken in the

*SOU - Statens Offentliga Utredningar.

**ERU - Expertgruppen for Regional Utredningsverksamhet.

1 Stockholm

I

Per thousand of mean population

1

Natural increase Net internal migration Net external migration

-

20

1

3 South Middle

n

I I

20

1

I I

4 South

Natural increase Net internal migration

I I

-

mean population

I I 1 I I

20 ^I

1

6 North Middle

Natural increase N e t internal migration h e r external migration

-

20

1

7 Lower North

Per thousand of mean population

8 Upper North

Natural increase Net internal rniaration

u

-

1970s when the growth of the metropolitan regions began t o lose momentum.

For Stockholm this meant a net loss of population in 1973 of approximately 1,300. This change in growth pattern is displayed in a reversed way in the northern regions, except for the Lower North region, which were experiencing an increase in population growth from the beginning of the 1970s.

2.1.1 FERTILITY

Table 2.1 gives total fertility rates for the county regions. The rates have been calculated on the basis of the number of births 2 years before and after each of the census years between 1950 and 1970 and for the single year 1974. The total fertility rates for county regions, to a limited extent only, depict the development of fertility at the national level. Over the entire 25-year period, fertility declined by about 20 percent in southern and central Sweden and by between 25 and 30 percent in the north. At thesame time the regional variation has been reduced as is seen from the standard deviations (unweighted) in the lower part of Table 2.1.

The analysis of regional variationsin fertility carried out at the Forecasting Institute for 1968- 1973 corroborates this general view. The regional pattern is fairly complicated. High-fertility regions are concentrated in the western part of the country. The Stockholm region is the most heterogeneous region in thls respect. The central part of the region (the Stockholm municipality) has the lowest fertility level in the country while 3 suburban municipalities in the region have the highest level. Fertility development during 1968-1973 varied through- out the regions. Those with a high fertiGty rate during the first part of the period showed a steeper fall latterly than those with a lower level initially. The analysis also confirms the previous conclusion that regional differences have been smoothed out t o some extent during recent years. Not only the level of fertility varied between regions, but also the structure of fertility. To the east the fer- tility of women over 30 became lower. On the other hand, the fertility of younger women was higher in the eastern part than in southwestern Sweden. In central Sweden the downward movement of childbirth into younger age groups became most pronounced. In the northem-most part of the country fertil- ity was still relatively high for women over 30, while fertility for women under 30 was low. The total numbers of births by age of mothers and by regions in 1974 are given in Appendix A. The age- and region-specific fertility rates are shown in Appendix B.

In a preliminary study with data from Jonkoping county, in the southern part of the country, a number of economic factors were analyzed with respect to their relation to a fertility measure (see Andersson and Holmberg 1978). In a regression analysis where all women were considered as one single group, the relation to economic variables tended t o be fairly weak. However, the very strong, positive relation with marital status and age, as explanatory variables, and the number of children may tend to obscure possible influences of socioeconomic

TABLE 2.1 Total fertility rate by region (x 1,000); census years 1950- 1970 and 1974.

Total fertility rates Region

Stockholm East Middle South Middle South West

North Middle Lower North Upper North Unweighted

average 2,358 2,232 2,352 1,935 1,834

Standard

deviation 209.94 90.45 144.67 101.48 95.39

Coefficient o f variation

(%) 8.9 4.1 6.2 5.2 5.20

' ~ o t a l fertility rates 1950 = 100.

SOURCE: Central Bureau o f Statistics.

variables on fertility as measured in the above way. Therefore, a new set of regression equations were estimated separately for all women and for married women distributed by age in 5-year age groups (20-24,

.

.

In this analysis age was found t o be an important factor in all age groups ex- cept for the highest. Family incomehad a negative effect upon fertility whenever it was included. Education too was a factor with negative influence on fertility.

The result of this preliminary analysis of regional variation in fertility pointed t o the importance of including differences in social variables between regional environments in further studies. Changes in background variables may have a considerable influence o n the future development of fertility within regions. In the present study specific fertility data are used for each region (Appendix A). The differences in fertility levels between the regions are seen in Table 2.1. The variations in the total fertility rate reflect, t o some extent, the variations accounted for in the study by the Forecasting Institute mentioned previously.

2.1.2 MORTALITY

Table 2.2 presents infant mortality rates for the county regions for 5-year periods (1 9 5 1-1 955,

. . .

TABLE 2.2 Infant mortality rates b y region; 5-year periods 195 1-55 t o 197 1 -

1975 (per 1,000).

Region 1951-55 1956-60 1961-65 1966-70 1971-75 Indef Stockholm

East Middle South Middle South West North Middle Lower North Upper North Unweighted

average 19.5 Standard

deviation 1.84 Coefficient of

variation (%) 9.4

'1nfant mortality rates 1951-1955 = 100.

years and the level for 197 1-1 975 is, on average, only 5 0 percent of the 195 1 -

1955 level. As in the case of fertility rates the regional variation of infant mor- tality has been reduced.

Table 2.3 presents a mortality index defined as the sum of the age-specific mortality rates for ages 0-79. The mortality index takes into account both infant and adult mortality. Regions in the south of Sweden appear t o have a slightly lower level of mortality than the rest of the country. The difference between the highest and the lowest level of mortality was approximately 1 4 percent in 1974 as compared to approximately 22 percent in 1950. According to this measure mortality has declined by almost 2 0 percent over the 25-year period.

There is n o apparent correlation between the levels of infant and adult mor- tality. The variations are in opposite directions for many regions. The rank correlation is, in fact, higher between the levels of fertility and infant mortality than between infant and adult mortality (0.38 and 0.14 respectively).

The regional variation in mortality has been studied at the Forecasting Institute and a report has been published.* According t o this investigation there exist differences in mortality that may have a considerable influence on regional population projections. A calculation of life tables for counties for the period 1966- 1970 revealed a significant difference in the mean expectation of life at birth. For males the lowest value was 70.57 years (Stockholm county) and the

*Regional dodlighet 1970-1975 (Regional Mortality 1970-1975). Information i Prognosfr%gor 1978:6 (Forecasting Information).

TABLE 2.3 Mortality indices (sum of age-specific mortality rates 0-79 years x 1,000) for regions; 1950, 1960, 1970, and 1974.

Region Stockholm East Middle South Middle South West North Middle Lower North Upper North Unweighted

average Standard

deviation Coefficient of

variation (%)

Mortality index

' ~ o r t a l i t ~ index 1950 = 100.

SOURCE: Hofsten and Lundstrom (1976). 1950-1974 Befolkningsfor'kdringar. Del 3 (Population Changes. Part 3).

highest value was 73.47 years (Kristianstad county in southern Sweden). A corresponding variation for females was a low of 75.79 years (Vi-mland county in the western inland areas and a high of 78.1 1 years (Kristianstad county).

The factors causing these variations have only been crudely analyzed. Pre- liminary analysis reveals the major factors t o be associated with industrializa- tion and urbanization. For a deeper analysis a more detailed regional system is required, e.g., the study carried out by the Forecasting Institute.

Most of the variation is leveled out when regions are aggregated, as in this study. Nevertheless, a certain amount of variation persists (Table 2.4). The fig- ures are calculated from the mortality schedule of a given region by the single- region life table program contained in the IIASA package (Willekens and Rogers 1978). A smaller range in the mean expectations is also affected b y the smooth- ing effect of migration. Migration can mean that newborn children spend only 30-40 years in their region of birth. According t o the hypothesis employed in the IIASA analyses, people who migrate immediately attain the mortality level in the region o f their destination. A more reasonable hypothesis would have to be formulated in terms of duration-of-stay effects.

TABLE 2.4 Expectation of life at birth, by sex and region; 1974.

Region Males Females

Stockholm East Middle South Middle South West North Middle Lower North Upper North

Total average

2.1.3 M I G R A T I O N

The age distribution of migration has a very distinct pattern common t o many countries (e.g., Rogers e t al. 1977). Apart from random fluctuations, the migra- tion rates shown in Figure 2.2 may be taken t o represent all counties in Sweden.

The IIASA population projection model also requires data on migration from a given region to each of the remaining regions. A gross migration matrix (GMM) is given in Table 2.5. Both men and women have a very similar distri- bution of migration over regions. The table also shows that a large proportion of all migration is concentrated in destinations that are geographically close t o the home region. The Stockholm region is an exception from this rule: migration from this region is distributed more evenly over the remaining regions. The two northern regions are at the opposite extreme, with over 3 0 percent of the migra- tion directed toward the Stockholm region and only 15 - 16 percent of the mi- gration going t o the nearest region.

There is a tendency for women moving from the two northern regions t o choose the highly urbanized Stockholm region as their migration destination.

Men seem to be more tied to regions close t o their home region. In the age- group 20-24 (Table 2.6) this tendency becomes more pronounced for both sexes. Young people in all regions, with few exceptions, seem to have a very strong propensity to move t o the Stockholm region.

2.2 Regional Disaggregation and Aggregation used in the IIASA Population Projections

The various administrative regions mentioned in Section 1.2 are not well suited for demographic analysis. The exception is the system of A-regions. These were constructed t o represent the boundaries of local labor markets. The A-regions are a f great importance for spatial population analysis because variations in

TABLE 2.5 Gross migration matrix (GMM). Percentage distribution. All ages.

Region of origin

Stockholm East Middle South Middle South West North Middle Lower North Upper North Region of

destination M F M F M F M F M F M F M F M F

Stockholm East Middle South Middle South West

North Middle Lower North Upper North

Total 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100 100

SOURCE: Calculated from raw data (Appendix A).

TABLE 2.6 Gross migration matrix. Percentage distribution. Ages 20-24 years.

Region of origin

Stockholm East Middle South Middle South West

Region of North Middle Lower North Upper North

destination M F M F M F M F M F M F M F M F

Stockholm East Middle South Middle South West North Middle Lower North Upper North

Total

SOURCE: Calculated from raw data (Appendix A).

-

,,, Female

0

0 ' 1 b 20 ' 3 0 40 50 60 70

I I 1 1 1 1 1 1 1 1 1 1

Age

FIGURE 2.2 Age-specific annual migration rates by sex;averages for the period 1968-1 973.

Source: Andersson and Holrnberg (1 978).

labor-market conditions are closely related t o migratory movements. Therefore, a regional system based on labor-market regions would be a valuable basis for the study of internal migration. On the other hand, such a large number of regions may be difficult t o handle, even with the use of a computer for analysis.

Several attempts have been made t o construct other regional systems more suitable for computer-based demographic analysis. These regional systems were designed t o be used for studying fertility, mortality, and migration.

In a study of regional variations in fertility in Sweden for the years 1968- 1973 a regional system of 11 7 fertility regions was constructed using munici- palities as the primary units. The following conditions were used in the delinea- tion of the fertility regions: contiguity, similarity of level of urbanization, sim- ilarity of economic structure and migration pattern, and similarity of levels of labor-market participation rates for women. Most municipalities were included in regions comprising 2 or 3 municipalities; 38 of the largest municipalities formed regions of their own. One of the leading principles in the construction of these regions has been t o obtain a sufficiently large population for the calculation of age-specific fertility rates.

Regional variations in mortality have been studied on several occasions in

Sweden. In a 1976 study attempts were made to construct mortality regions on the basis of counties. Four distinct clusters could be separated for males: the first cluster comprised counties in southern and western Sweden, the second counties in central Sweden, the third included the metropolitan counties and the two northernmost counties, and the fourth, which only comprised three counties, was in the southeastern part of the country. The regional variation in female mortality was less pronounced and the clusters were more homogeneous.

There was an aggregation t o 5 clusters: the first cluster was formed by the two largest metropolitan counties, Stockholm and Gothenburg the second included the two northern counties, the third was formed by the counties in southern Sweden, and the fourth and fifth clusters were formed from the counties in central Sweden.

The most important conclusion t o be drawn from the study was that there exists a distinct pattern in the regional variation in mortality. This conclusion has led to the implementation of a more comprehensive study of regional mor- tality, including the construction of a consistent regional system for mortality studies. In this study, which was carried out at the National Central Bureau of Statistics, municipalities and A-regions were used as primary units.

In the study,* which was part of the series Forecasting Information pro- duced by the Forecasting Institute at the National Central Bureau of Statistics, it was revealed that no subdivision into a smaller number of regions, as in the IIASA studies, would be sufficient to describe the regional mortality variation.

The third attempt to construct a regional system was initiated in a study of internal migration by the authors of this report (Andersson and Holmberg 1978). Preliminary estimates of migration equations appeared to have a distinct spatial structure. It was, therefore, considered necessary to determine whether the country could be regarded as homogeneous with respect t o the determinants of migration. A subdivision of the country into different clusters in whlch the migration behavior is homogeneous and separated from the rest of the country may be primarily motivated by the upgrowth of spatial regions with very high information density. Because of this, a number of cluster studies were carried out on the A-regions. The purpose of the studies was t o see whether regions with a similar distribution of out-migration and in-migration were linked to- gether in any apparent spatial pattern. As a criterion of the difference between A-regions with respect to the destination of their migration a x2-measure was used. This measure expresses the deviation between observed and expected frequencies for each pair of rows in the migration matrix. The expected fre- quencies are calculated on the basis of a hypothesis of a uniform distribution with fixed marginal distributions.

A cluster analysis is designed to aggregate elements for which there are observations on a number of variables. The observations in this case were each of the 70 A-regions regarded as elements and in-migration from the remaining

*Regional dodlighet 1970-1975 (Regional Mortality 1970-1975). Information i Prognosfr%gor 1978:6 (Forecasting Information).

69 regions to a given region as a vector of variables; the total in-migration to the given region was regarded as the 70th variable. The results are shown in Figure 2.3 where seven clusters of A-regions are distinguished.

The country can be crudely subdivided into two main regions. The northern region can be divided further into two subregions and the southern region into four subregions.

Since the symmetric property of the matrix representing net migration is not preserved, when gross migration is considered, we may reasonably assume that the transposed migration matrix may also be of interest to study. In this case the regions are clustered with respect to the destination of their out- migration. The result is shown in Figure 2.4 and only some minor differences in comparison with the above analysis are revealed. The two main regions are divided into six subregions, of which two are to be found in the northern main region.

The purpose of the cluster analysis was t o establish whether regional mi- gration in the country shows any specific patterns and if it is possible to form large contiguous regions that are homogeneous with respect to regional migra- tion patterns. A close correspondence was found t o exist between the systems of regions obtained by considering similarities in the migration flows both by origin and by destination.

The cluster analysis was used only as an indication as to how the final regional system should be constructed. The first step was to separate the three large metropolitan A-regions: Stockholm, Gothenburg, and Malmo. Furthermore it was judged necessary to represent separately the inland and coastal areas of the two northern regions. The final result was 13 so-called M-regions.

The regional system employed in the present study is similar to the system of M-regions. Table 2.7 gives an overview of the relationship between the three systems.

2.3 Observations o f Aggregation Errors Due t o Limitation o f the Number o f Regions

The IIASA models for population projections are normally adapted to a small number of contiguous and rather large regions. This can cause important errors of aggregation in the sense that large intraregional variations in mortality, fer- tility, and mobility are disguised in macroaggregates. Our studies of regional mobility patterns in Sweden have shown that such errors of aggregation are not so serious in the case of migration. There is, however, reason for caution in the cases of mortality and fertility. As indicated in Figures 2.5 and 2.6 high- mortality and high-fertility locations are scattered all over the country without any easily detected and smooth distance-decay effects as there are for migration.

The aggregation problems are especially pronounced in the Upper North region. There subareas of very high fertility are mixed with ones at the other extreme of the fertility distribution.

Region 1 Region 2 Region 3 Region 4 Region 5 Region 6 Region 7

FIGURE 2.3 Functional regional aggregation on the basis of origin of in-migrants. Source:

Calculated from raw data. 195 1-1 960 Ekfolkningsrorelsen; 196 1-1 966 Folkmangdens fdiandringar; 1967-1975 Befohiigsfdirindringar. Del3 (Population Changes. Part 3).

Reg~on 1 Region 2 Region 3 Region 4 Region 5 Region 6

FIGURE 2.4 Functional regional aggregation on the basis of destination of out-migrants.

Source: Den inrikes omflyttningen 1968-1973: storlek, monster och flyttningsavstgnd (Internal migration in Sweden 1968-1973: size, pattern and migration distance); Informa- tion i PrognosfrPgor 1974:9 (Forecasting Information).

TABLE 2.7 Comparison of three different regional systems: the M-regions, the county regions and the CMS Study regions.

CMS Study County M-region

region region

1 Stockholm 2 East Middle 3 South Middle 4 South 5 West 6 North Middle 7 Lower North 8 Upper North

1 AB 2 CDETU 3 FGHI 4 KLM 5 NOPR

8 ACBD

(county and A-region)

1 Stockholm A-region AB except A2 -

5 Sm%land FGHI

6 Southern Sweden KLM except A28

7 Western Sweden NOPR except A33 in 0 county and A43 in P county

8 Southwest inland S and A46 in T county 1 1 Middle North inland Z and A61 in Y county 12 Upper North inland AC, BD

1 3 Upper North coastal area AC, BD 2 Gothenburg A-region

3 Malmo A-region 4 ~ s t e r ~ i j t l a n d E

9 Central Sweden TUW and A5-A7 in D county except A46 in T county

1 0 Middle North coastal area CXY and A2 in AB county except A61 in Y county

SOURCE: Hofsten and Lundstrom (1976). 1974 Befokningsforkdringar. Del 3 (Population Changes.

Part 3).

NOTE: For explanation of symbols see Figure 1.1, cf. also Figure 1.2.

3 MULTIREGIONAL POPULATION ANALYSIS 3.1 Introduction

Population models are based o n a number of assumptions, such as constant fer- tility and mortality and the absence of external migration. Classical demographic analysis has been oriented to the development of populations over time. The spatial aspect was not included. In contrast, spatial and also social mobility has been a major focus of interest for economists, geographers, and sociolo- gists. During recent years demographers have been increasingly interested in formulating multiregional versions of available demographic models.

Originally, Lotka (1 907) derived his population growth model in continuous terms. The discrete formulation was suggested by Bernadelli (1941) and in a more elaborate way by Leslie (1945). By introducing a matrix operator, demo- graphic projections may be carried o u t easily. Although theoretically the results of the continuous and discrete models are analogous to each other, the discrete approach has certain advantages, especially when large computers can be used for the calculations.

Formulated as a discrete matrix model, the one-region population growth

FIGURE 2.5 Natural increase of population per 1,000 of mean population (1975). Source:

Befolkningsforbdringar 1975. Del3 (Population Changes 1975. Part 3).

FIGURE 2.6 Live births per 1,000 women of ages 15-44 years (1975). Source: Befolk- ningsforiindringar 1975. Del 3 (Population Changes 1975. Part 3).

model is easily expanded to a large number of regions; the number of regions is mainly limited by the capacity of the computer. The assumptions underlying this multiregional model are similar to those on which the one-region model is based: age-specific rates of fertility, mortality, and migration are fixed and the population is closed t o external migration. Migration between the regions within the model is, however, permitted.

The basic mathematics of multiregional demographic growth have been elaborated at IIASA and are contained in computer programs t o provide users with a ready tool for population analysis. These programs, and the underlying theoretical analysis, are published elsewhere (Willekens and Rogers 1978).

The IIASA projection models have been used for the projections for Sweden, presented in this report.

The following sections deal explicitly with the characteristics of the multi- regional population change. First, the multiregional life table for Sweden is presented; fertility and mobility analysis is then performed using the multi- regional demographic theory, and finally the results of a multiregional popula- tion projection for Sweden are discussed. These projections are camed out with constant schedules of fertility, mortality, and migration. The schedules of age- specific rates are given in Appendix B.

3.2 The Multiregional Life Table

The multiregional life table shows the combined effect of mortality and mobility on survival of individuals in a set of regions. All biometric functions of the single-region table are thus given a multiregional equivalent. In order t o illustrate the multiregional population system, multiregional life-table functions are presented below. Table 3.1 gives the probabilities of survival from birth to age 20 by region. These probabilities tell us what proportion of a given birth cohort will survive 20 years. Information is given not only about total survival but also about where the person survives.

Regional differences in survival probabilities are negligible. There is, how- ever, a much larger variation in the proportion surviving in the region of birth.

The largest value, 0.77, is recorded for the West region and the smallest, 0.63, is for Stockholm. Table 3.1 also gives some idea about the variations in direc- tional migrations between regions. Cohorts born in the northern regions generally have a higher probability of survivingin the Stockholm region than cohorts born in the South and West regions (Table 3.1, first row).

Expectations of life at birth and at age 20 are shown in Tables 3.2 and 3.3.

The complete table for expectation of life is given in Appendix C. The calcula- tion of regional life tables is based on the hypothesis that the mortality rate which applies to an individual is determined by his region of residence. This means that a person who moves is exposed to the mortality rate prevailing in his new region of residence. This leads t o a number of interesting consequences.

The mortality rate which applies to an individual is determined not only by his

TABLE 3.1 Probabilities o f survival from birth to exact age 20 by region. Total population; 1974.

Region of residence at age 20 Stockholm East Middle South Middle South West North Middle Lower North Upper North

Total

Region of birth Stockholm 0.63146 0.1 1226 0.03533 0.04349 0.04670 0.05661 0.03102 0.02878

East Middle 0.09820 0.63476 0.03768 0.041 38 0.06055 0.06364 0.02123 0.02664

South Middle 0.05513 0.06791 0.64697 0.08166 0.08744 0.02205 0.00722 0.01167

South 0.04357 0.03752 0.0499 1 0.76456 0.05572 0.01713 0.00734 0.00974

West North Middle Lower North Upper North

SOURCE: From computer output of model applications.